Corante

About this Author
DBL%20Hendrix%20small.png College chemistry, 1983

Derek Lowe The 2002 Model

Dbl%20new%20portrait%20B%26W.png After 10 years of blogging. . .

Derek Lowe, an Arkansan by birth, got his BA from Hendrix College and his PhD in organic chemistry from Duke before spending time in Germany on a Humboldt Fellowship on his post-doc. He's worked for several major pharmaceutical companies since 1989 on drug discovery projects against schizophrenia, Alzheimer's, diabetes, osteoporosis and other diseases. To contact Derek email him directly: derekb.lowe@gmail.com Twitter: Dereklowe

Chemistry and Drug Data: Drugbank
Emolecules
ChemSpider
Chempedia Lab
Synthetic Pages
Organic Chemistry Portal
PubChem
Not Voodoo
DailyMed
Druglib
Clinicaltrials.gov

Chemistry and Pharma Blogs:
Org Prep Daily
The Haystack
Kilomentor
A New Merck, Reviewed
Liberal Arts Chemistry
Electron Pusher
All Things Metathesis
C&E News Blogs
Chemiotics II
Chemical Space
Noel O'Blog
In Vivo Blog
Terra Sigilatta
BBSRC/Douglas Kell
ChemBark
Realizations in Biostatistics
Chemjobber
Pharmalot
ChemSpider Blog
Pharmagossip
Med-Chemist
Organic Chem - Education & Industry
Pharma Strategy Blog
No Name No Slogan
Practical Fragments
SimBioSys
The Curious Wavefunction
Natural Product Man
Fragment Literature
Chemistry World Blog
Synthetic Nature
Chemistry Blog
Synthesizing Ideas
Business|Bytes|Genes|Molecules
Eye on FDA
Chemical Forums
Depth-First
Symyx Blog
Sceptical Chymist
Lamentations on Chemistry
Computational Organic Chemistry
Mining Drugs
Henry Rzepa


Science Blogs and News:
Bad Science
The Loom
Uncertain Principles
Fierce Biotech
Blogs for Industry
Omics! Omics!
Young Female Scientist
Notional Slurry
Nobel Intent
SciTech Daily
Science Blog
FuturePundit
Aetiology
Gene Expression (I)
Gene Expression (II)
Sciencebase
Pharyngula
Adventures in Ethics and Science
Transterrestrial Musings
Slashdot Science
Cosmic Variance
Biology News Net


Medical Blogs
DB's Medical Rants
Science-Based Medicine
GruntDoc
Respectful Insolence
Diabetes Mine


Economics and Business
Marginal Revolution
The Volokh Conspiracy
Knowledge Problem


Politics / Current Events
Virginia Postrel
Instapundit
Belmont Club
Mickey Kaus


Belles Lettres
Uncouth Reflections
Arts and Letters Daily
In the Pipeline: Don't miss Derek Lowe's excellent commentary on drug discovery and the pharma industry in general at In the Pipeline

In the Pipeline

« C&E News: No Smiley Face This Time | Main | How Much Fraud, As Opposed to Plain Old Error? »

November 6, 2012

Time For An Election Post

Email This Entry

Posted by Derek

Well, every other web site in the US will be going on about the election today, and with good reason. So I'll put up a quick post of my own, because I'll be glued to the returns tonight myself. Along the way, I've been teaching my children how to interpret them, which makes them, I'd say, among the few local middle-schoolers who know that Florida has two different poll closing time (the panhandle's on CST), to wait for Pennsylvania because Philadelphia's ballots always seem to drag in last, and that on a country-by-county basis, Ohio looks like something that Dr. Frankenstein assembled on his day off. My father was an election commissioner when I was growing up back in Arkansas, and it left a mark. I've found, though, that a background in the Arkansas politics of that era has served me well, not least in making me difficult to shock when it comes to the behavior of politicians during (and after) elections.

So what do I think is going to happen? Well, I'm used to seeing raw biological assay data, so when I see people giving probabilities of political victory with figures to the right of the decimal place, I just smile. All I'll say is that I think it's going to be a close call for whoever wins, and that anyone (on either side) who's confident it won't be needs to get out more. Were I a betting man (perish the thought), I'd put some money down on a Romney upset, because I think you could get some good odds, thanks to Nate Silver. But we'll see - tonight, or in the morning, or (God help us all) even later than that.

Comments (54) + TrackBacks (0) | Category: Current Events


COMMENTS

1. Curious Wavefunction on November 6, 2012 8:45 AM writes...

I guess modeling of elections is at least one notch up in terms of complexity relative to modeling of protein-drug interactions.

Permalink to Comment

2. Virgil on November 6, 2012 8:46 AM writes...

My bet is Obama wins the electoral college while Romney wins the popular vote. Hopefully this won't end up in a Gore-V-Bush-esque lawsuit free for all.

Permalink to Comment

3. luysii on November 6, 2012 8:54 AM writes...

My wife and I have been happily canceling each other's vote since '64. Our marriage hasn't suffered. Differences of opinion is what makes horse races, which is why I find the near unanimity of opinion where I live so stultifying. They're all for diversity, as long as it doesn't include ideas. For a rant on this subject see

https://luysii.wordpress.com/2012/10/15/the-fruits-of-liberal-governance/

Permalink to Comment

4. Nekekami on November 6, 2012 8:56 AM writes...

As a Swede and a miltiary veteran: I'd rather have the predictable normal nastiness of Obama than the crazy, random and unpredictable nastiness of Romney.....

Also... I'm amazed at how the US election committees still think that electronic voting machines is the way to go, when computer scientists, programmers, electronic engineers etc all come out and say that with the proper procedures, pen and paper is still the hardest to rig...

As a programmer and software engineer myself, I'm against the electronic voting machines.

Permalink to Comment

5. Hap on November 6, 2012 9:06 AM writes...

I don't think Romney winning would be that much of an upset. I don't like his (party's) politics, but he could at least be competent, which would be good in any case. If the difference in polls is the margin of error, nothing (other than a blowout) should be surprising.

I'm really hoping that someone wins outright, though - I don't think going a vote going to the SC will do anyone any good and "may the best lawyers win" is not a good motto for a just electoral system.

Permalink to Comment

6. johnnyboy on November 6, 2012 9:15 AM writes...

@5: I think you meant "may the most partisan supreme court win".

Permalink to Comment

7. Chris D on November 6, 2012 9:24 AM writes...

The election probabilities aren't anything magic: they're just averaging the polls for each state and then applying that to the scenarios required to win the Electoral College. As Nate Silver pointed out, unless the polls are systematically statistically biased against Romney, the electoral math makes a victory for him very difficult.

The popular vote probabilities are national poll averages, but the nationwide popular vote is a sideshow: that fact that it's nearly tied is utterly irrelevant except as a rhetorical device after the election is over.

If you think the numbers are systematically wrong, shouldn't you have a theory as to why?

Permalink to Comment

8. Hap on November 6, 2012 9:24 AM writes...

That too.

It's getting harder for me to determine what the worst-case scenario is.

Permalink to Comment

9. rhodium on November 6, 2012 9:27 AM writes...

My guess is that pollsters like Silver really know that three significant figures is nonsense. However, the less mathematically literate will look at the numbers and think this data is the best around. He does have a brand to sell after all. People will remember who got the winner right and a few of the swing state predictions, but not the actual projected vote percentages.

Permalink to Comment

10. Ed on November 6, 2012 9:29 AM writes...

I only found out today (because I checked) that there are in fact more than two candidates! My vote goes to Gary Johnson.

Permalink to Comment

11. LeeH on November 6, 2012 9:43 AM writes...

Derek

I'll give you 2:1 odds against Romney (I'll take Obama).

Permalink to Comment

12. Derek Lowe on November 6, 2012 10:05 AM writes...

Lee, I believe that Nate Silver's probabilities would call for 8:1 or more, depending on how much "vig" the house takes. He's a confident guy.

Permalink to Comment

13. My 0.02 on November 6, 2012 11:01 AM writes...

INTRADE has Obama at 72.4% - that is about 3:1 odds for Obama. I'll take INTRADE's number over Nate's - although both point to same conclusion.

Permalink to Comment

14. Chris D on November 6, 2012 11:07 AM writes...

You guys know Silver predicted 50 out of 51 correctly in 2008, right? (He missed Indiana but was correct on D.C.)

I think y'all are confusing Silver with either a pollster or a pundit. He's a giant nerd applying sabermetrics--which has been stunningly successful in baseball--to politics. And he's been successful at it already.

Permalink to Comment

15. Wile E. Coyote, Genius on November 6, 2012 11:21 AM writes...

How did Silver do in 2010? Good, but not as spectacular as the 2008 predictions. I think we can, over the long term, expect a regression of his performance to the mean (like most stock pickers perform in financial markets). Is his model superior? Maybe, but time will tell. My guess is that turnout this year will be more like 2010 than 2008. I wouldn't bet on election results for 2012 based on Silver alone.

Permalink to Comment

16. Mo on November 6, 2012 11:22 AM writes...

Considering how close these presidential races usually are I would have expected that you needed more than one election for the results to be statically significant.

Permalink to Comment

17. Nekekami on November 6, 2012 11:44 AM writes...

@16

Indeed.... Once or even twice can easily be good guesswork or even luck.

Permalink to Comment

18. Rev. Howard Furst on November 6, 2012 11:59 AM writes...

Then there's this sort of confounding factor in polling analysis - fraudulent internal polls, probably intended at the time to stampede the yahoos and gin up the rubes, but "corrected" (scapegoated) in time to not be quite so embarrassing:

The Romney campaign now saying internals attributed to their pollster Neil Newhouse showing Mitt up in Ohio, tied in WI, PA "are incorrect"

Permalink to Comment

19. My 0.02 on November 6, 2012 12:00 PM writes...

@15,

I won't equate polling with stock picking. What Nate does is data mining/analysis. I truly believe that stock market is totally random.

Permalink to Comment

20. Question 2 on November 6, 2012 12:17 PM writes...

MA residents: make sure to educate yourselves on shockingly flawed Question 2 "Death with Dignity". Of course it's in insurance companies interest to end people's lives early!

Permalink to Comment

21. Wile E. Coyote on November 6, 2012 12:17 PM writes...

I think there is some science to stock picking (i.e. stocks that outperform the market) based on an understanding of business fundamentals. Warren Buffet has it figure out. This is more true than the analysis of Silver, which is data mining of polls that may or may not have been based on appropriate sampling assumptions and may or may not have been appropriately executed. Silver doesn't actually do the polling; he looks at results of polls and other information and makes predictions.

Permalink to Comment

22. Anonymous on November 6, 2012 12:37 PM writes...

@20: I am sorry you seem to have bought into the right's fearmongering that physician-assisted suicide is nothing but Big Brother death squads. There are many people who are suffering and have no hope of improving, and would like to have some degree of choice in how they spend their final days. In any case, glad I got to put that in writing in my vote this morning rather than just debate it on a blog.

Permalink to Comment

23. TOSG on November 6, 2012 12:47 PM writes...

I'm a bit disappointed to see this sort of "forget data, I'm just gonna go with my gut!" type thinking from you, Derek. Do you disagree with the reliability of the input data or with Silver's methodology, or do you just not like his conclusion?

Permalink to Comment

24. Vanzetti on November 6, 2012 1:00 PM writes...

@20

>>>I'm a bit disappointed to see this sort of "forget data, I'm just gonna go with my gut!" type thinking from you, Derek.

He is a drug maker. It goes with the job.

Permalink to Comment

25. Anonymous on November 6, 2012 1:05 PM writes...

@20: I am sorry you seem to have bought into the right's fearmongering that physician-assisted suicide is nothing but Big Brother death squads. There are many people who are suffering and have no hope of improving, and would like to have some degree of choice in how they spend their final days. In any case, glad I got to put that in writing in my vote this morning rather than just debate it on a blog.

Permalink to Comment

26. derek lowe on November 6, 2012 1:05 PM writes...

TOSG, that wasn't what I meant to imply. Silver's a smart guy, but his model rests on several assumptions which I don't think are are easy to assign as the (spurious) precision of the final number might make some people think. Turnout is one of the biggest - what's the composition of the electorate going to be? As for his method of dealing with such data, the problem is, the inner workings of Silver's model are proprietary. No one else, as far as I know, gets to look under the hood.

So with those in mind - the uncertainty of the input data, and the uncertainty of the machinery used to process it - the treatment of Silver and 538 as Scientific Truth is a bit annoying. Here's a further look that you might find interesting: http://www2.macleans.ca/2012/11/04/tarnished-silver-assessing-the-new-king-of-stats/

Permalink to Comment

27. Anonymous on November 6, 2012 1:06 PM writes...

@20: I am sorry you seem to have bought into the right's fearmongering that physician-assisted suicide is nothing but Big Brother death squads. There are many people who are suffering and have no hope of improving, and would like to have some degree of choice in how they spend their final days. In any case, glad I got to put that in writing in my vote this morning rather than just debate it on a blog.

Permalink to Comment

28. Kismet on November 6, 2012 1:29 PM writes...

Derek:
All betting markets* and all poll aggregators** consistently show an Obama advantage and something like 60-90%+ probabilities for an Obama win. How can anyone believe the race is close? Of course, this does not exclude a Romney win but it's far from a tossup, isn't it?

* Intrade and betfair
** Sam Wang, electoral-vote.com, Drew Linzer, Pollster, Talking Points (as per Krugman)

Permalink to Comment

29. TFox17 on November 6, 2012 1:35 PM writes...

@26: Derek, as I understand it, likely voter considerations are largely taken into account by the individual pollsters, not the Silver model. His methodology is described here: http://fivethirtyeight.blogs.nytimes.com/methodology/. I don't think anyone claims its Scientific Truth, it's just a betting line. As a critique, you link to a Colby Cosh opinion piece; Colby is an amusing writer, but I note that he's not making any predictions at all.

Permalink to Comment

30. luysii on November 6, 2012 1:48 PM writes...

#20, #27 -- This retired MD voted against physician assisted suicide because of the very high likelihood of abuse. Death is the ultimate economy. Look at what government is already doing in the name of raising revenue -- lotteries, casinos. It's just grinding the face of the poor in modern guise.

State government also has a definite economic interest in keeping people smoking and drinking. Do you really think they'll want to keep you alive if you have a very expensive, very chronic and nonfatal illness. I don't.

The best example of this sort of thing is what happened to the institutionalized mentally ill. Mental hospitals were closed and patients were discharged to essentially nonexistent community care. It was the perfect storm politically. The right didn't want to pay for care. The left was all about autonomy. Talk to the next homeless person you meet.

Permalink to Comment

31. Question 2 on November 6, 2012 3:01 PM writes...

#25/30: many reasons to be against:
- Doctors strongly oppose assisted suicide
- Advocates for the disabled strongly oppose assisted suicide
- Question 2 does not require family notification.
- Question 2 does not require a patient to consult with a psychiatrist or palliative care expert before receiving the lethal prescription.
- Assisted suicide would weaken efforts to expand and improve palliative care.
- Complex issues like assisted suicide should be decided in a legislative process rather than a ballot initiative.
- Terminal diagnoses are often wrong.

Permalink to Comment

32. LeeH on November 6, 2012 3:03 PM writes...

Derek

Alright. 3 to 1.

Permalink to Comment

33. Doug Steinman on November 6, 2012 3:25 PM writes...

I'm not sure that it really matters who wins tonight. I think THE WHO said it best,"Meet the new boss. Same as the old boss".

Permalink to Comment

34. Fred on November 6, 2012 3:38 PM writes...

I have go with the Dems, every time, because:

1) the absolute, number 1 goals of the GOP since Reagan are tax cuts for the rich and super-lax regulation. Tax cuts for the rich DO NOT seem to trickle down to yours truly, and though I'd love to see someone reform the FDA so they are not such a destructive band of drug-killing "nervous Nellies", we have all seen what weak regulation regarding the environmental and the banking system can do.
2) The Dems are about expanding civil liberties; the GOP caters to the extreme, "anti-everything" crowd-- anti marriage equality, anti-woman, anti DREAM Act, anti-education, pro-nonsense (Creationism, et al). I want everybody to get as fair a chance in life as practical.

The reasonable Republicans (Lugar, etc) have been shown the door. Romney may have been moderately moderate in the day, but he's owned by Rove now, and all the other dudes feeding his machine (Obama may be a bit beholden to Steven Spielberg and tons of small, individual donors, but that I can easily live with).

My Humble Opinion.

I think Nate is right to predict a win, but I expect at least one state to "pull a Florida"-- so I hope Obama picks up enough EV's so he doesn't need it.

Permalink to Comment

35. Anonymous on November 6, 2012 4:11 PM writes...

my friend says if you have a bookie, you can get obama at -400 and romney at +320. ie 1:4 for obama, 3.2-1 for romney


I remember it was 2:3 for obama 6 months ago

Permalink to Comment

36. Henning Makholm on November 6, 2012 4:12 PM writes...

@19: If the stock market is Totally Random, then you should be able to make a bundle by large-scale "volatility pumping". Which isn't a paradox, because the money would come from the poor schmucks who think they have an inkling which way it is going. (If everyone were following that strategy it would stop being profitable, of course -- because it acts like negative feedback on the market, tending to make prices stay at wherever they are).

Permalink to Comment

37. McChemist on November 6, 2012 7:13 PM writes...

Kismet/28:

The problem is that all those poll aggregators are all pretty much looking at the same underlying data - so if there is some systemic bias in the underlying data, all those aggregators would be incorrect. And a bunch of people at Intrade are relying on those, so it's pretty natural that Intrade's odds would sort of line up with the poll aggregators.

Then again, the odds at Intrade for Romney (~30%) are a lot better than 538 (8%) or Sam Wang (pretty much zero). That difference is probably the odds the market puts on the polls being biased against Romney. If there's no bias in the data (or the polls are biased FOR Romney), then you've obviously got an Obama win.

Permalink to Comment

38. Ex-Pat on November 6, 2012 7:13 PM writes...

This election is as dysfunctional as it can get. Every state doing voting differently? Elections should be run by a none partisan federal organization, with crayon and paper! Just look at Elections Canada, everyone votes by placing a mark in a circle next to a name on a piece of paper. No results before all polls are closed and NO LAWYERS!

Permalink to Comment

39. McChemist on November 7, 2012 12:35 AM writes...

Welp, it appears the underlying data the poll aggregators were looking at was dead on. Nate Silver wins again (well, and Obama too).

Permalink to Comment

40. Kismet on November 7, 2012 6:50 AM writes...

Systematic bias might be an issue, but we don't (didn't) know which candidate it would favour AFAIUnderstand. Also, some of this certainly should cancel out because pollsters, aggregators and betting markets have different though overlapping methodology--

Permalink to Comment

41. Philip on November 7, 2012 9:37 AM writes...

@Nekekami (4), I slightly disagree with your no electronic voting post.

Electronic voting machines are the best way to fill out a paper ballot. Pen and paper can be more difficult to fill out and read (see Florida in 2000) than voting machines. The computer can make sure you have voted in a logical way. For example it can catch two or more votes for president. For some elections such as school board you may be able to vote for multiple people, the machines can make sure you have voted for the correct number (it must include a line that states that you do not want to use all or any of your votes).

When you have filled out the ballot with the electronic voting machine, it counts the votes and prints a paper ballot. The paper ballot then goes into an optical reader and is counted again. The paper ballot is the official ballot. Both electronic voting machine and optical reader counts are checked. If there is enough of a difference, the paper ballots are recounted.

I am going to skip other details, such as when a voter makes a mistake, but I think you get the picture.

@McChemist (39), you beat me to that post. As a left wing talking head said last night, Nate Silver was the big winner.

Permalink to Comment

42. LeeH on November 7, 2012 9:47 AM writes...

Actually, we modelers are the winners. The closeness of the predictions with reality, despite a reasonably wide range of methods (OK, with a lot of actual "experimental" poll data), demonstrates how predictive methods can be useful.

Now if only molecules behaved as predictably as humans...

Permalink to Comment

43. Nekekami on November 7, 2012 5:24 PM writes...

@41

Speaking as someone who's working with secure and/or heavily audited software, the problem is NOT as trivial as you try to make it out. It's a pain in the ass to make secure voting software that can leave a good audit trail, YET which can not tie a vote to an individual voter in any way whatsoever, even if that method is by checking the timestamp and comparing with observers in the voting station.

People hold up slot-machines etc as an example, but that's trivial compared to the voting machines. Those only have to have a verified pseudo-random number generator and proper probability weighting, while voting machines need what I mentioned before, yet it also needs to log and identify any poll workers, any accesses to the hardware, not just the software. Nevermind the fact that you can't trust the hardware either, because once an attacker has the hardware in their hands, it's game over.

Every new poll needs to have the code recertified, to see that everything works as intended, that there are no unintended escapes etc etc. And formal verification outside of extremely small codebases is insanely complex and time-consuming. And you have to trust that the programmers are good enough at spotting seemingly innocuous code that is in fact malicious(cross-reference the Underhanded Coding contests to see just how nasty that can be.). Even strict languages like ADA, Erlang and similar have been susceptible to it, and most software developers churned out by unis and colleges simply aren't good enough to spot it).

In short, it's easier to rig a couple of electronic voting machines, even if they are designed properly, than it is to rig a properly setup paper ballot with machine-assisted hand-counting(pro-tip, ballot boxes never leave the polling station until the counting is done, and there are always at least 3 observers from various parties etc overseeing the counting. MUCH harder to tamper with that, since you need to get many more people involved in the tampering, and thus greater risk of blowing the entire thing)

Permalink to Comment

44. Philip on November 7, 2012 6:04 PM writes...

@Nekekami (43), All that you said is true, and why I insist that the electronic voting machines produce a human and machine readable paper ballot that is the official ballot. The electronic voting machines main function it to make sure that the voter has filled out the ballot in a proper way. Its secondary function is to allow for a quick unofficial counting. With today's technology we should not be waiting hours for vote counts.

Permalink to Comment

45. Bell4 on November 8, 2012 8:52 AM writes...

So Nate Silver seems to be batting 1.000 this time around. Pretty good outcome in most leagues. Derek, care to comment?

I have to say, it is disappointing to this long-time reader of your superb med chem/pharma commentary to see you outsource your credibility to right-wing hacks when politics comes up.

Permalink to Comment

46. Derek Lowe on November 8, 2012 10:55 AM writes...

Summed up well here, Bell4, on that well-known nest of right-wing hacks known as Slate:

http://www.slate.com/articles/news_and_politics/politics/2012/11/nate_silver_obama_s_big_win_doesn_t_mean_the_538_guru_is_an_electoral_genius.html

Permalink to Comment

47. Jacob on November 8, 2012 11:04 AM writes...

Derek, that article is one of the dumber ever posted on one of the dumber websites on the web. For shame.

Permalink to Comment

48. Bell4 on November 8, 2012 12:14 PM writes...

I did read the Slate article, Derek, thanks. It's a pretty dreadful piece - now that Silver's methodology again succeeded for the presidential race (one of the few with the extensive data sets needed to make high quality calls), this guy claims it was obvious from the beginning. Sheesh. And having a few non-wingnuts sniping at Silver doesn't change the reality that he has overwhelmingly been a target of vituperation from the right. This kind of selective citation is not your proudest moment.

Permalink to Comment

49. Nekekami on November 8, 2012 12:19 PM writes...

@44

If they can't fill out a paper ballot correctly with a pen, the machine won't help them much. The very checking code you mention will be a nightmare to verify also.

Also, why the needless stress about the counting taking hours? It hurts more to have all that needless stress than it hurts to have to wait for a few hours.

Permalink to Comment

50. McChemist on November 11, 2012 9:31 PM writes...

Wait, Derek, I'm confused. Pre-election, you're saying the odds are skewed towards Obama since Nate Silver is (presumably incorrectly, in your view) proclaiming him to be the obvious favorite, and that might not be correct.

Now you're linking to articles saying that Silver wasn't saying anything we didn't already know, and we should have all realized how easily Obama was going to win this election based on pre-election polls (on which Silver based most of his methodology).

It seems like you're being a little inconsistent here.

Permalink to Comment

51. Derek Lowe on November 12, 2012 7:38 AM writes...

What I was trying to say before the election what that I thought that Silver's turnout model might be flawed (that is, based too much on 2008). As it happened, the turnout was different than 2008, but both parties went down, with the Republicans going down more.

I linked to the Slate article largely because I wanted to show that it was possible to criticize Silver from other than a hands-over-the-ears right-wing perspective. But it's true that the Slate article comes at him from a "But this is what the polls said all along" perspective, where I was wondering if the polls were right in the first place.

I think what gets on my nerves is that all the triumphalism on the part of Silver's supporters, the "You can't argue with math!" stuff. I've seen too much of what can happen with modeling and forecasting in my own field to treat anyone (or any method) as an oracle.

Permalink to Comment

52. LeeH on November 12, 2012 2:24 PM writes...

Derek

Based on Silver's (somewhat cryptic) description of his method, it doesn't seem that turnout (likely voter adjustment) plays a very large role. It does seem that (along with the polls themselves) there is a much larger component that tries to compensate somewhat for previous prediction accuracy, as well as extrapolating based on trend.

http://fivethirtyeight.blogs.nytimes.com/methodology/

I do agree with you that this is not a simple "math" issue. However, it is certainly a "weight of evidence" issue. Multiple samplings, from multiple sources, over multiple time points.

This debate reminds me of the global warming "controversy". While there is no absolute "proof" of global warming, there are multiple measurements made by the majority of scientists over multiple time points. Much like the aggregators that, almost unanimously, came to the same conclusion about the outcome of the election.

Permalink to Comment

53. Bell4 on November 12, 2012 5:29 PM writes...

I actually wasn't a cheerleader for Silver as such. He only came to my attention because of the tidal wave of scorn he elicited from the Right, a fair amount of it ad hominem. Their denials of the possibility that his methods could lead to accurate election predictions was reminiscent of their treatment of numerous 'controversies' in the realm of science, well-known to you and your readers. Your remarks echoed the sense, although not the tone, of their invective. When the election results came out so strongly in his favor, I expected nothing from the political pundits in the way of mea culpas. Your usual generosity and sense of fair play, however, led me to hope that a gracious remark would be forthcoming, regardless of your disappointment with the electoral outcome and your unease (which I share) about the actual scientific merit of his techniques.

Permalink to Comment

54. McChemist on November 12, 2012 7:33 PM writes...

Derek,

Understood. Given your skepticism of modeling in general, you might actually like Nate Silver's book. It's a pretty nice review of forecasting in different fields, and it's written with a skeptic's frame of mind.

Given that the state of election forecasting in the media still often falls along the lines of "random guy predicts victory for whichever candidate has political views closest to his own", it's nice to see more empirical methods get their time in the spotlight. Getting two straight presidential elections pretty much correct doesn't make Nate Silver a genius, but it does make him quite a bit better than lot of others in his current field.

Permalink to Comment

POST A COMMENT




Remember Me?



EMAIL THIS ENTRY TO A FRIEND

Email this entry to:

Your email address:

Message (optional):




RELATED ENTRIES
XKCD on Protein Folding
The 2014 Chemistry Nobel: Beating the Diffraction Limit
German Pharma, Or What's Left of It
Sunesis Fails with Vosaroxin
A New Way to Estimate a Compound's Chances?
Meinwald Honored
Molecular Biology Turns Into Chemistry
Speaking at Northeastern